OpenAI is limiting access to its new cybersecurity tool, GPT-5.5 Cyber, rolling it out only to “critical cyber defenders” — the same approach CEO Sam Altman previously criticized rival Anthropic for taking with its own cybersecurity tool, Mythos.
Altman announced on X on Thursday, April 30, 2026, that Cyber would begin reaching select users within days. Those seeking access must submit an application on OpenAI’s website detailing their credentials and intended use.
According to the application, Cyber is designed to assist with penetration testing, vulnerability identification and exploitation, and malware reverse engineering — functioning as a toolkit to help organizations find security gaps and stress-test their defenses. The concern driving the restricted rollout is that the tool could be misused by malicious actors if made broadly available.
The announcement carries an ironic edge. When Anthropic restricted access to Mythos under similar reasoning, Altman publicly dismissed the move as “fear-based marketing.” Other critics echoed that view, arguing Anthropic’s warnings were overblown. Adding to the awkwardness, an unauthorized group reportedly gained access to Mythos anyway.
OpenAI says it plans to expand Cyber’s availability over time by consulting with the U.S. government and identifying additional users with legitimate cybersecurity credentials.
The parallel between the two companies’ approaches suggests that concerns about dual-use AI security tools may be pushing developers toward similar access controls, regardless of how those controls are characterized publicly.
Source: TechCrunch